Written by Teacup Lab.
This is the fourth article in our SkincAIr series, which explores the contextual realities of implementing a healthcare app designed to support the diagnosis of neglected tropical diseases affecting the skin (skin NTDs).
In July 2026, our research team travelled to Senegal to complete the third and final round of discovery fieldwork for the SkincAIr project. We visited health facilities across Dakar and Touba, spoke with frontline health workers, dermatologists, community health promoters, and policymakers from the national NTD programme.
Read our previous posts on the limitations of desk research and our fieldwork in Kenya and Ethiopia.
Cultural adaptation in digital product design can mean many things. At the surface level, it means adjusting the interface for a new market: language, date formats, local conventions. That kind of work matters. But it operates at a different layer from what fieldwork keeps showing us.
When we spend time in the field, inside health facilities, observing clinical workflows, talking to people who will actually use what we are building, what we tend to find is not a list of surface-level adjustments. We find that some of the assumptions built into the core logic of the product need to be revisited.
In Senegal, º: where the user journey begins, what the proposed tool must understand, who needs to be part of the system, and what value the product must add to practices that already work.
The user does not always arrive at the beginning
Most digital health tools are designed around a defined clinical encounter. The patient arrives, the clinician examines, the tool supports a decision. The workflow is bounded and relatively linear.
What we experiment across fieldwork is that by the time a patient sits down in front of a clinician, there is already a history. The neighbour who recognised the rash. The pharmacist who suggested something over the counter. The traditional practitioner who recommended a remedy. Sometimes all of these, in sequence, over weeks or months. For conditions like mycetoma, which can go undiagnosed for many years, this journey can be very long.
The clinician is consequently not starting from scratch. They are inheriting a history, and that history is clinically significant. What was applied before the consultation may have altered the lesion, changed its colour, its borders, its texture. A dermatologist we spoke with put it plainly: history-taking is not a preliminary step before diagnosis. It is diagnosis.
For SkincAIr, this shifts how we think about what needs to happen before any AI analysis takes place. An image of a lesion whose appearance has been altered by a previous remedy, without the context of what that remedy was, can point in a misleading direction. A “previous treatments” field is part of the answer. But the deeper question is about what the tool’s entry point should be, and what kind of clinical context needs to be captured first.
What fieldwork offers here is the ability to see where the real process begins, which is often earlier, and messier, than the workflow diagram suggests.
Clinical categories do not always map onto local ones
Digital health tools are necessarily built around clinical categories. Diseases have names, and those names organise the diagnostic logic, training data, and interface. What is less obvious is how often those categories do not align with the categories users and patients actually think in.
For instance, dermatologists and other healthcare actors explained why patients often arrive having already tried treatment for a different condition. The explanation began with a word: ndoxum siti.
It is a Wolof-French expression derived from the word for syphilis, a disease once associated with highly recognisable skin manifestations. Over time, the expression became detached from its original meaning and evolved into a catch-all for almost any weeping or crusting skin condition. Participants described ndoxum siti as a term that community members and traditional practitioners may use for conditions as different as scabies, leprosy, mycetoma, and fungal infections. One term, many diseases: one remedy applied to them all.
The practical consequence for diagnosis is significant and this is not a problem that a language option solves. It asks something more structural: how does the tool bridge between the community terminology a patient arrives with and the clinical differentiation the tool is trying to support?
The path to formal care runs through other people first
In both Kenya and Ethiopia, we had noted that patients often consult traditional practitioners before reaching a clinic. In Senegal, we found a more formalised model of collaboration: in several districts, health teams had spent years deliberately building working relationships with traditional practitioners and marabouts (Muslim religious or spiritual guides who may hold significant community influence), not to displace their role, but to connect it to formal care.
One community health coordinator described what this looked like in practice. His team had mapped some marabouts and traditional practitioners in the district, trained them to recognise warning signs, and given them physical referral slips to pass on to patients. They had created a shared WhatsApp group that included the district physician alongside the practitioners.
“Most of them are highly cooperative now. Some of them even accompany their patients to the hospital in person.”
A Ministry of Health coordinator offered the underlying logic: “You need the marabout’s endorsement, and people will adhere.”
What this approach reflects is a pragmatic observation about how trust operates in communities where religious and traditional authority carries significant weight. Teams that had invested in working with these actors had found that their influence could become an asset.
For SkincAIr, this has implications for who the relevant stakeholders are in a Senegalese deployment, and for the role that community health actors, including formally recognised roles like the Badiénou Gokh (community outreach volunteers embedded in neighbourhoods), might play in early case identification. A simple interface for identifying suspected cases and referring patients designed for this layer could extend the tool’s reach to moments earlier in the journey than any clinic-based solution can reach.
In this sense, fieldwork allows to reveal the people who shape adoption but are absent from the initial clinical workflow.
There is already something doing the job
Ultimately, one of the more useful things fieldwork does is show you what already exists in the space you are designing into.
Across the facilities we visited, in both urban and more peripheral settings, health professionals were already using WhatsApp to seek informal advice from colleagues and specialists about skin cases. They shared photos of lesions and clinical context, using back-and-forth exchanges to discuss possible diagnoses and next steps. When a presentation was typical, they could advise the clinician remotely. However, when it was ambiguous, they recommended that the patient be seen in person.
This is already a functioning form of informal clinician-to-clinician case consultation. It is fast, draws on professional networks people already trust, and is so embedded in daily practice that most clinicians do not describe it as a digital tool.
SkincAIr would therefore enter an established practice rather than introduce a new one. Its success depends on whether it offers enough value to justify changing a familiar workflow. A formal tool can offer what WhatsApp does not reliably provide: structured case records, traceability, governance, and images linked to a patient file. But those benefits must be clear and immediate, or the informal system will remain the rational choice.
What this means for SkincAIr
These findings point to the same conclusion: cultural adaptation shapes where the workflow begins, what information the product needs, who may influence or use it, and which existing practices it must complement.
A literature review might identify the role of marabouts or describe delayed care-seeking. A questionnaire might capture WhatsApp use. But fieldwork revealed what these realities mean for SkincAIr’s design: the product must account for previous treatments, bridge local and clinical categories, work through trusted referral networks, and offer clear value alongside existing clinician-to-clinician consultation practices.
This final discovery round does not give us a finished answer. It gives us a more grounded set of design decisions to take back to the people who helped shape them.
This article marks the end of the discovery phase of our From Fieldwork to Design series. Prototype validation visits across Kenya, Ethiopia, and Senegal are planned for late 2026.
If you want to always stay updated about our project, follow us on LinkedIn, Facebook, X and Bluesky!